Paper No. 33 Determinants of Regional Entry and Exit in Industrial Sectors Kristina Nyström

CESIS Electronic Working Paper Series Paper No. 33 Determinants of Regional Entry and Exit in Industrial Sectors Kristina Nyström (JIBS, CESIS) May 2005 The Royal Institute of technology Centre of Excellence for Science and Innovation Studies http://www.infra.kth.se/cesis Corresponding author: kristina.nystrom@jibs.hj.se

Determinants of Regional Entry and Exit in Industrial Sectors Kristina Nyström Jönköping International Business School P.O. Box 1026 SE-551 11 Jönköping Sweden E-mail: kristina.nystrom@jibs.hj.se Abstract: Recent empirical research by, for example, Audretsch and Fritsch (1999) and Armington and Acs, (2002) shows that regional determinants of new firm formation differs between industries. It has also been suggested that a large part of the regional variation of new firm formation can be explained by differences in industrial structure. This paper reinvestigates the regional determinants of entry and exit considering these findings. The empirical analysis is performed using data on Swedish firm entry and exit rates for 1997-2001. It is shown that on average about 0.5 to 2.7 percent of the regional variation in entry and exit rates remains to be explained, after controlling for differences in industrial structure, but that there is substantial regional variation. A majority of the firms in the 47 industries investigated are sensitive to unobserved regional characteristics, such as regional policy when deciding to enter or exit a particular region. Agglomeration and the size structure in the particular industry and region are factors that are found to influence entry and exit rates in almost all industries. Keywords: Entry, exit, industry structure, regions JEL classification code: L1, R12-1 -

1. Introduction An important decision for entrepreneurs who consider starting a new firm is to decide were to locate their business. For a firm already in business the location at a certain point in time is fixed but does not have to persist in the future and therefore there can still be regional factors influencing the decision to exit or not. Today there is an extensive literature on the determinants of entry and exit of firms in a regional context providing valuable empirical evidence from different time periods, regions and industrial sectors. However most studies tend to focus on the manufacturing industry and do not recognize differences between industrial sectors. It is well known that there are substantial differences in entry and exit rates between industrial sectors, especially between the manufacturing and service sectors. (See for example Nyström 2001.) Some recent papers, by for example, Audretsch and Fritsch (1999) and Armington and Acs, (2002) also find that regional factors determining new firm formation differ between industries. Armington and Acs, (2002) also note that most studies on regional determinants of entry use variables such as unemployment rate, population density and availability of financing as explanatory variables. The development of theories of new economic geography and endogenous growth theories implies that such aspects should be incorporated in the analysis. Fritsch (1997) shows in his paper that the number of regional new firm start-ups clearly depends on the industry structure in the region. He finds that more than half of the new firm startup s can be explained by industry differences. If the empirical analysis does not consider that relationship he emphasize that there is a risk that the empirical results reveal regional industrial differences rather than differences in regional factors determining entry and exit of firms. Therefore it is necessary to add new empirical evidence on these issues incorporating both regional and industrial variations in the analysis of entry and exit of firms in a regional perspective. In addition to incorporating additional explanatory variables and focusing on differences between industries it is also interesting to note that most of the literature in this area focuses only on entry of firms. This paper will focus on both entry and exit and we will see if the regional factors usually used to explain entry also are important determinants of exit. The purpose of the paper is twofold; firstly it intends to investigate how much of the Swedish regional differences in entry and exit rates during the 1990s can be explained by industry - 2 -

structure. Secondly it will also study if regional determinants of entry and exit rates vary between industrial sectors. Methodologically the paper contributes to the already existing empirical literature by applying a panel data approach, that takes unobserved regional specific effects into account.the paper is organised as follows: section two presents the literature on determinants of entry and exit in a regional perspective. The third section presents the data, variables and methods used to empirically study these relationships. Section four presents the regional variations in entry and exit rate and compares them with entry and exit rates adjusted for regional industrial structure. Section five provides the results from panel data regressions on determinants of entry and exit rates in different industries. Finally conclusions and suggestions for future research are presented. 2. Determinants of regional entry and exit There is a vast literature trying to explain the regional differences in entry and exit rates. These studies usually identify three major categories of factors influencing spatial differences; local demand factors, the supply of founders and the policy environment. (Keeble et. al. (1993), Johnson and Parker (1996)). These three categories are defined with starting point from different perspectives on the process of new firm formation and firm births, but they can all be further illustrated by thinking about what is important for an individual who is thinking about starting a firm. A potential entrant wants the business to be profitable. The local demand factors reflect the market potential for the new firms. Are there any potential customers and can they afford to buy the good supplied by the entrants? The supply of founders perspective focuses on who the individuals that start new firms are. What other opportunities than to start a new firm do they have and what knowledge do the individuals in particular regions possess? The policy environment reflects, for example, what kind of support, both in terms of financial support and knowledge support an individual that is planning to start a new firm can get from local authorities. In some regions there might be policies available trying to keep exit rates low. The problem with policy environment factors in an empirical analysis is that such information is difficult incorporate in a quantitative analysis. Of these three categories of factors this paper will focus on local demand factors and supply of founders, but the policy environment will be - 3 -

implicitly included in the analysis since we use a panel data approach that account for unobserved regional effects. In addition to the local demand factors and supply of founders factors we will also consider agglomeration as a factor influencing a firms decision to locate in or exit a particular region. The agglomeration approach emphasize that firms can benefit from locating close to each other. Based on Marshall s (1920) pioneering work the theories of agglomeration and co-location emphasize that firms can benefit from locating close to each other in several ways. The purpose of such cooperation s is to decrease the costs and that can be done in different ways. They can share some of the fixed cost with other firms, or manage to together with other firm negotiate increased accessibility and lower prices of local inputs. A co-location of firms can more easily attract skilled worker to the region, which is beneficial to all firms in the industry (Krugman, 1991). The proximity to other firms may also increase the possibility of knowledge spill-overs and innovation. Note that there are not always only positive effects of locating close to other firms. If too many firms locate close to each other it might cause increasing wages and increased input prices when they compete for the same resources. The different factors included in the local demand factors, supply of founders and agglomeration will be further described below in order to find variables that can be included in the empirical analysis. However, we have to elaborate a bit further on how and why it is important to take industrial structure into account when we analyse the regional determinants of entry and exit rates. The importance of the different factors influencing entry and exit rate can, as mentioned before be expected to differ between industries. The influence of the different factors can be expected to depend on, for example, the cost structure in the production, the elasticity of demand with respect to income changes, the amount of investment needed to start a new firm, if its worthwhile to locate close to other firms, or what kind of knowledge that is required in a specific industry. A geographical application of the product life cycle theory, as suggested by, for example, Hirsch, (1967) provides us with a theoretical framework, explaining that the needs of a firm differ between different stages in the product life cycle. In the initial stages capital and skilled labor are needed to develop the product. Hence, a firm producing such a good prefers to be located in a region with good access to such factors. As the product matures and new entering firms compete with offering a lower price of the product the firm needs to lower its input cost. The choice of - 4 -

location will therefore be based on the access to cheap input factors and the choice of location could therefore be a region with lower labor costs. 2.1. Local demand factors The decision where to locate can be expected to be influenced by the opportunities, in terms of demand, offered by various regions. A major determinants of demand is the number of potential customers i.e the firms and inhabitants in the region. and their incomes. In this section we focus on the demand from private persons. The demand and interrelation to other firms is further discussed in the agglomeration section. Population: Since the firms in a region provide the inhabitants with goods and services demanded, both the size of the population and changes in population can be of importance. If firms tend to locate in regions with large population, represents a self-reinforcing effect. When new firms start or move to a region it becomes even more attractive for other firms to locate there since the size of the region increases further (Krugman,1998). The opposite self-reinforcing effect causing exits might of course occur if the population is small and decreasing. Income: Another factor influencing demand in a specific region is obviously the incomes earned in the region. Also in this case both the income level and changes in the income level can be expected to influence entry and exit rates. Increased levels of incomes increases demand, but also the access to capital that a potential entrant needs in order to start a firm. (Reynolds, 1994) In the discussion of the impact of incomes it is important to remember that the incomes also constitute labor costs for the firms. A high income level might therefore also deter entry in industries that are sensitive to high labor costs. 2.2 Supply of founders The category supply of founders reflects the individual s incentives to start a new firm or close down an unprofitable firm. Unemployment: The theoretical arguments suggest that high rates of unemployment can both have positive and negative effects on entry rates. For example, (Storey, 1991) discusses that high unemployment rates can cause higher entry rates, since it forces unemployed workers to start their own companies as an alternative to unemployment. The alternative hypothesis would be that there is a negative relationship between unemployment and new firm formation. Such an - 5 -

approach is based on the view of unemployment as a measure of the general economic situation, and therefore high and increased levels of unemployment decreases demand and cause less entry and more exit. The relationship between unemployment and entry of firms has been investigated in several empirical studies, with sometimes contradictory results. Carree (2002) summarizes the empirical evidence and concludes that in time series analysis of the relationship between unemployment and new firm formation, the empirical studies often find a positive relationship, whereas cross-section studies find a negative relationship. Education: The educational level of the population in a region is expected to influence entry and exit of firms due to the fact that the probability of starting a new firm is higher for well educated people (e.g. Evans and Leighton, 1990). As earlier mentioned it might also be more attractive for firms producing products in the initial stages of the product life cycle to locate in a region with access to a well educated workforce. On the other hand, a high education level might not be desirable by firms producing mature products according to the product life cycle theory. Firm size: The market structure in a specific industry is expected to influence entry and exit rates. Industries with in general small-scale firms tend to have higher entry rates since it is more common with spin-offs from small firms. The higher spin-off rates in small firms are explained by the fact that the employees in small firms discover market opportunities more easily and that they have knowledge about how to operate a small firm. (Johnson and Parker, 1996) The size structure also reflects that it is easier to start a new firm in an industry were economies of scale are not that important, i.e., in industries were entry barriers are low. If entry barriers are high exit barriers can also be expected to be high (e.g Shapiro and Kehmani, 1987). 2.3 Agglomeration effects The agglomeration effects perspective tries to focus on the advantages of being located close to other firms. Spillovers and co-operation between firm can occur both between firms in the same industry, and between firms operating in different industries. The literature distinguish between localization economies and urbanization economies, (e.g. Hoover 1937, Ohlin, 1933) where localization economies refer to the advantages of being located close to other firms in a particular industry. Urbanization economies on the other hand are the external effects associated with the size and density of a region. - 6 -

Localization economies: If there are many other firms in the same region in the same industry it might attract potential entrants to the region. The benefits in terms of cost reductions and spillovers can also make incumbent firms more successful than firms that are not co-located. If the co-located firms are more profitable, they have better chances to survive and help each other if, for example, demand decreases temporarily. Therefore we expect exit rates to be lower in agglomerated industries. In relation to the discussion on localization economies it is however important to note that the proximity to other firms is not always advantageous for a firm. A cluster of firms working in the same industry still means that competition among the firms heavily influence their behavior and profitability. Urbanisation economies. Firms may also benefit from locating close to other firms even though they do not belong to the same industry. Urbanization economies in terms of benefits associated with location in a large region is already discussed and captured by including a population variable. In addition to locating in a large region firms may experience positive external effects from locating in a dense area. Lower transport cost and closeness to suppliers and customers reduces cost and improves the quality of the good or service produced. In connection to the urbanization economies it is important to note that at a certain regional size and density diseconomies of agglomeration may exist. (see for example Richardsson, 1995) Such diseconomies may occur due to, for example traffic congestion, increased land rents or increased labor costs. 3 Data, method and description of variables 3.1 Data and variables The data used in the empirical analysis are collected by Statistics Sweden and consist of firm level data where the firms are classified as belonging to different industries according to the Standard Industrial Classification (SIC) system on the 5-digit level. The data consist of information regarding the financial situation for enterprises in the corporate sector. All industries included in the SIC-classification system, except Financial intermediation (SIC-code 65-67) Real - 7 -

estate activities (SIC-code 70) and Activities of membership organizations (SIC-code 91) are included in the dataset. Data for 1996 to 2001 are available, which makes it possible to compute entry and exit rates for the five years 1997 to 2001. The data was aggregated to two-digit SIClevel. Industries with SIC-code 11 (Extraction of crude petroleum and natural gas) 13, (Mining of metal ores) 16, (Manufacture of tobacco products) and 41 (Collection, purification and distribution of water) were not included in the empirical analysis, since each of these industries consists of less than 20 firms and therefore have very few entry and exits. This means that 47 different industries remain to be included in the analysis. For firms with more than 50 employees the data are based on a survey conducted by Statistics Sweden and for firms with less than 50 employees the data are based on other administrative sources. This data collection method means that all Swedish firms (except for firms in the sectors mentioned above) are included in the dataset. The firms are coded in a way that makes it possible to identify when each individual firm enters or exits. The total number of observations in the dataset for each year is between 200.000 and 300.000. The data include information from jointstock companies, cooperatives, partnerships, limited partnerships, associations and some foundations. The dataset includes financial information from the profit and loss account and balance sheet as well as some basic data such as the number of employees and value added. In order to only take firms with real economic activity into account, firms with no employees and firms that reported no sales were not included in the dataset. After removing these non-active firms almost 200.000 firms remained for each year. In addition to the information in the dataset described above, data from Statistics Sweden on income, unemployment, education and population were used. The regional classification used in this paper was developed by NUTEK (1998). In the classification, based on commuting patterns, Sweden is divided into 81 local labour market regions. Two alternative approaches of computing entry and exit rates are usually used; the ecological approach or the labour market approach. The ecological approach relates the number of entering or exiting firms to the number of already existing firms in a specific industry, whereas the labour market approach relates the number of entering or exiting firms to the number of employees in the industry (Armington and Acs 2002) In this paper the ecological approach is used, since it is - 8 -

the industry structure perspective that is the main interest in this paper. The definitions of the variables are provided in detail in Table 1. Table 1: Description of variables Variable Definition E Entry rate: The number of entering firms in industry i, region r at time t divided by r, i, t the number of firms in industry i region r at time t. X Exit rate: The number of exiting firms in industry i, region r at time t divided by the r, i, t number of firms in industry i region r at time t. Pop Population: The population in region r at time t r, t Pop Population change: Population in region r at time t minus population in region r at r, t time t-1 divided by the population in region r at time t Inc Income: Total income from employment and business in region r at time t.(in fixed r, t prices using harmonized CPI 1996 as deflator.) Inc Income change: The total income from employment and business in region r at time r, t t minus total income from employment and business in region r at time t-1 divided by the total income from employment and business in region r at time t.(in fixed prices using harmonized CPI 1996 as deflator) Unem Unemployment: The Number of unemployed aged 16-64 in region r at time t r, t (including persons in unemployment programs)/population aged 16-64 in region r. Unem Unemployment change: Unemployment rate in region r at time t minus r, t unemployment rate in region r at time t-1. Edu Number of employees with a university degree in region r /number of employees in r, t region r. Size Firm size: A concentration measure summing the squared individual firms share of r, i, t the employment in industry i and region r. Loc Localisation economies: Number of firms in industry i, region r divided by the r, i, t population in region r. Urb Urbanisation economies: The number of firms in region r divided by the population r, t in region r. Dum Industry dummy variable i 3.2 Adjusting entry and exit rates for industrial structure One way to take regional differences between industries into account is to do a type of shift-share analysis by subtracting the country average number of entrants from the number of real entrants for a specific industry and region (Fritsch 1997). The adjusted entry rate E r, i ( adjusted) for region r and industry i at a specific time period t, is then calculated as follows: - 9 -

E ENTRY r, i, t r = 1 r, i, t ( adjusted) = 81 Nr, i, t (1) 81 ENTRY r = 1 N r, i, t r, i, t where ENTRY r, i, t is the number of entering firms in region r and industry i at time t, and N r i, t, is the number of firms in region r and industry i at time t. When we subtract the average entry rate for the whole country in an industry, from the actual industry entry rate in the region we get the adjusted entry rate. The adjusted exit rate ( X r, i, t ( adjusted) ) is calculated correspondingly. Note that the adjusted entry and exit rates can have negative signs if the actual entry rate in a specific region is lower than the average entry or exit rate of the industry in the country. 3.3 Panel data method As mentioned earlier the results of empirical studies on determinants on entry and exit rates tend to vary between industries and regions. This obviously calls for a panel data approach that makes it possible to incorporate unobserved regional differences when analyzing which factors influence entry and exit rates. A one-way error-component fixed effect panel data model is appropriate since we expect there to be unobserved characteristics of each region that needs to be accounted for in the analysis. 1 The analysis will be made at three different level of aggregation which is necessary in order to facilitate a deeper understanding of what determines regional entry and exit. Below these three different levels of aggregation will be named A, B and C when they are specified as models to be estimated. A) Aggregate regional level E r, t = α + β 1 Pop r, t + β 2 Pop r, t + β 3 Inc r, t + β 4 Inc r, t + β 5 Unem r, t β 7 Edu r, t + β 8 Size r, i, t + β 9 Urb r, t + µ r + v r, t + β 6 Unem r, t + (2) 1 An alternative approach could be to estimate a two-way fixed effect model. Such a model also includes unobserved time specific effects. In our econometric model we have several time-specific variables reflecting the state of the business cycle and therefore the inclusion of time effects would cause problems with multicollinearity. - 10 -

B) Industry and regional level E r, i, t = α + φ1pop r, t + φ2 Pop r,, t + φ3inc r, t + φ4 Inc r, t + φ5unem r, t + φ6 Unem r, t + φ7edu r, t + φ8size r, i, t + βφ9loc r, i, t + φ10urb r, t + φ11dum i +... φ57dum i + λr + ω r, t (3) C) Individual industry level In this case equation 4 is estimate for each of the 47 industries. E r, i, t = α + δ1pop r, t + δ 2 Pop r,, t + δ 3Inc r, t + δ 4 Inc r, t δ 7Edu r, t + δ8size r, i, t + δ 9Loc r, i, t + δ10urb r, t + ε r + γ r, t + δ5unem r, t + δ 6 Unem r, t + (4) In the panel data approach µ λ and ε r r r denotes unobservable regional specific effects, β,φ and δ are parameters to be estimated, Since we assume a one-way fixed effect model µ λ and r r ε r are fixed parameters to be estimated and v r, t ω and r, t γ r, t respectively are assumed to be independent and identically distributed with zero mean and σ v 2 and σ variance. The 2 ω explanatory variables are assumed to be independent of and t. (Baltagi 2001) v r, t ω and r, t γ r, t respectively for all r In order to check whether the fixed effect model is the appropriate model the Hausman specification test can be used. This test is based on the previously mentioned assumption that the explanatory variables are independent of v r, t ω and γ r, t r, t in the fixed effect models. The null hypothesis in the Hausman test is that we have a random effects model and the test statistic is 2 k distributed as χ where k is the number of explanatory variables. In the alternative random effects model mean and σ 2 µ µ, λ and ε r r r are assumed to be independent and identically distributed with zero 2 and σ variance. µ λ and ε r r r are assumed to be independent of λ γ r,t in each equation respectively. (Baltagi, 2001). v r, t ω and r, t - 11 -

4 Regional variations in entry and exit rates To get a picture of the entry and exit rates in the different Swedish regions the average entry and exit rates, using the ecological approach are computed. These regional entry and exit rates can be compared with the adjusted entry rate described in equation 1. Therefore the average adjusted entry and exit rates 2 for each region is calculated. These entry and exit rates are presented in Figure 1-4, but before we analyze these patterns it is interesting to look at the general Swedish pattern regarding the unadjusted and adjusted entry and exit rates. If we sum the average adjusted entry rates over all regions and divide it by the number of regions we get a country average. 3 The results from these computations are presented in Table 2. The table shows that if we do not control for industrial structure the entry rate is between 9.7 and 11.3 percent during the five different years but if we control the entry rates for industrial structure the average adjusted entry rate is only between -1.2 and -2.7 percent. When the same computations are performed regarding exit rates the unadjusted exit rate range between 9.2 and 11. 7 percent, but if we adjust for industry structure the exit rates varies only between -0.5 and -1.0 percent during the five years. Even though the empirical analysis in this case is on a aggregated country level it is obvious that much of the regional variation in entry and exit rates can be explained by the industrial structure. It is also worth mentioning that even though the average figures for the adjusted entry and exit rates are quite low, the standard deviation and minimum and maximum values show that the adjusted entry and exit rates vary quite a lot across regions. It is also interesting to observe that both the adjusted entry and exit rates have negative means, which indicate that a majority of the regions have lower entry and exit rates compared to the country average, given their industrial structure. This means that there is a skewness in the distribution of the adjusted entry and exit rates among the regions. Appendix A provides an example of how the distribution looks like for adjusted entry rates in the year 2000. 47 E 2 Regional Average adjusted entry rate r, i, t ( adjusted) = i= 1 47 81 Regional Average adjusted entry rate ) 3 Country average adjusted entry rate= r = 1 81-12 -

Table 2: Mean, minimum and maximum values of entry and exit rates 1997-2000 1997 1998 1999 2000 2001 Entry rate Mean 0.101 0.113 0.097 0.113 0.100 Minimum 0.059 0.045 0.039 0.056 0.033 Maximum 0.158 0.178 0.149 0.186 0.182 Std.Dev. 0.019 0.023 0.019 0.024 0.022 Entry rate adjusted Mean -0.012-0.027-0.016-0.022-0.020 Minimum -0.054-0.095-0.076-0.079-0.087 Maximum 0.045 0.038 0.034 0.051 0.062 Std.Dev. 0.019 0.023 0.018 0.024 0.022 Exit rate Mean 0.117 0.102 0.111 0.092 0.102 Minimum 0.033 0.066 0.075 0.047 0.056 Maximum 0.199 0.155 0.145 0.143 0.156 Std.Dev. 0.022 0.016 0.014 0.018 0.012 Exit rate adjusted Mean -0.008-0.005-0.007-0.010-0.010 Minimum -0.092-0.041-0.043-0.054-0.056 Maximum 0.074 0.049 0.026 0.041 0.044 Std.Dev. 0.022 0.016 0.014 0.018 0.019 Figure 1 and 2 gives a deeper understanding of the regional differences in entry rates by showing the regional average entry and adjusted entry rates for 1997-2001. The four different colors represent quartiles of the entry rates in the regions. Some regions in the inner parts of northern Sweden have the lowest entry rates whereas the two largest labor markets in Sweden; Stockholm and Gothenburg have among the highest entry rates. The construction of the adjusted entry rate variable suggests that a region with a positive adjusted entry rate is a region with higher entry rate than expected, taking the industrial structure of the region into account. Regions in the southwest of Sweden have the lowest adjusted entry rates. Note that Stockholm which had among the highest entry rates, now have among the lowest adjusted entry rates, and one reason for why Stockholm have high entry rates but low adjusted entry rates is that industries that on average tend to have high entry rates, such as many service industries, is located in Stockholm. If we compare Figure 1 and 2 we can also see that the problems with low entry rates in the inner part of northern Sweden in some regions can be explained by their industrial structure, since they perform somewhat better in terms of adjusted entry rates. - 13 -

Figure 1: Average entry rates 1997-2000 in 81 Swedish regions. Average entry rates 1997-2001 Entry rates (%) 7.5-10.2 10.2-12 12-13.6 13.6-17.4-14 -

Figure 2: Average adjusted entry rates 1997-2000 in 81 Swedish regions. Average adjusted entry rates 1997-2001 Adjusted entry rates (%) -10.3 - -5.1-5.1 - -3.8-3.8 - -1.6-1.6-2.2-15 -

Figure 3 and 4 compare the unadjusted exit rates with the adjusted exit rates. In these figures we can see that many of the regions that have the highest exit rates are located in the northern parts of Sweden. Stockholm and Gothenburg regions also have rather high exit rates. In the southern part, on the contrary, exit rates are low in many regions. The this case the construction of the adjusted exit rate imply that a positive adjusted exit rate means that we have a higher exit rate than expected in a particular region. Many of the regions with positive adjusted exit rates are situated in the northern part of Sweden. Therefore we can conclude that these regions are facing some severe problems with exiting firms, since they do not only have high exit rates but also higher exit rates than the country average exit rates in these industries. - 16 -

Figure 3 Average exit rates 1997-2000 in 81 Swedish Average exit rate 1997-2001 Exit rates (%) 8.2-10 10-10.8 10.8-12 12-15 gions.

Figure 4: Average adjusted exit rates 1997-2000 in 81 Swedish regions. Average adjusted exit rate 1997-2001 Adjusted exit rate (%) -6.7 - -0.9-0.9-0 0-1.1 1.1-4.7-2 -

5. Determinants of regional entry and exit rates in different sectors. In this section the results after estimating the regressions specified in equation 2, 3, and 4 are presented. The estimations are corrected for hetroscedaticity using White s hetroscedaticty consistent matrix (Greene, 2003). Appendix B provides measures of correlation between the different variables used in the regressions. This correlation matrix shows that the we don not expect any severe problems with multicolliniearity. Tables 3 and 4 presents the results of the empirical analysis when the aggregate regional data is used. Table 3 presents the results of estimating equation 2 were entry rates are the dependent variable. Since the value calculated according to the Hausman test statistics are higher then the critical value, the Hausman specification test suggests that we should choose the fixed effect model instead of the random effects model. A chi-square test also supports the choice of a fixed effect model with region specific effects instead of an OLS-estimation. In the table the results from the OLS estimation is also presented in order to be able to compare with earlier empirical research and alternative model specifications. In appendix B the regional specific effects estimated for each of the 81 local labour market regions are presented. From the fixed effect estimation we conclude that the most important local demand factor is the income level and changes in incomes. The income level has a negative impact on entry rate and a reasonable explanation to that is that, as mentioned earlier, the wage that the employees earns is also a cost for the potential entrant. Therefore some potential entrants do not want to locate in regions with high wage-levels. On the other hand if the income level increases it attracts new entrants that want to capture the additional demand in the region. In the category supply of founders both unemployment and unemployment change have significant coefficients influencing entry rates. The level of unemployment in a region lovers entry rates whereas an increase in the unemployment rate increases entry rates even though the magnitude of the effect is are small. A reasonable explanation would be that the level of unemployment is a general measure of the state of the economy and the business cycle, whereas the change in unemployment rate actually measures the dynamic aspect of what alternatives an employees that have lost their job have. The estimation result also shows that agglomeration effects are of great importance. The presence of - 3 -